Claude Skill MartBrowse skillsQuick linesLearn by videoTerminal guideWhat is a Skill?
 
← Back to list

Deep Research (Multi-Agent Research Orchestration)

Splits a research goal into parallel sub-goals, runs each in a headless `claude -p` subprocess, and aggregates everything into a polished standalone report file.

AutomationAdvanced★ 1,632⑂ 245AI score 9/10Last updated: Aug 13, 2026

What it does

  • Triggers on requests like "deep research", "wide research", or systematic investigation. It first clarifies the goal, gathers real samples via search/scraping, then presents a plan and waits for your explicit go-ahead.
  • Decomposes the goal into parallelizable sub-goals (link lists, topic clusters, time slices) and generates prompt files plus a dispatcher script under .research/<name>/.
  • For 3+ sub-goals it launches headless claude -p workers (8 in parallel by default) with per-task --allowedTools, timeouts (5–15 min), exit-code capture and tailable logs.
  • Aggregates child outputs into a draft, redesigns the outline, then iterates chapter by chapter instead of one-shot rewriting, delivering the final report as a file rather than a chat dump.
  • Includes prompt-injection defenses (all scraped content is untrusted data), raw-data caching, retry/failure isolation, and a pre-delivery self-check checklist.

Who it's for

  • Analysts, PMs and engineers who regularly run multi-hour competitive, market or technology research.
  • Anyone batch-processing dozens of links or datasets into a long-form, citation-backed report.
  • Claude Code CLI users, ideally with firecrawl or exa MCP configured (falls back to WebSearch/WebFetch).

Examples

  1. "Deep research the AI coding assistant market" → 12 sub-goals by vendor, pricing and features run in parallel, ending in an insight-driven long report file.
  2. Hand it 50 conference session URLs and ask for per-session summaries plus trend synthesis → batched with GNU Parallel, then merged.
  3. Investigate three years of regulatory changes split by year, producing a timeline where every claim carries an inline source link.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/feiskyer/claude-code-settings/HEAD/skills/deep-research/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the skills/deep-research folder from the GitHub repo feiskyer/claude-code-settings into my ~/.claude/skills/feiskyer-deep-research/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/feiskyer/claude-code-settings.git /tmp/ccs && mkdir -p ~/.claude/skills && cp -r /tmp/ccs/skills/deep-research ~/.claude/skills/ && rm -rf /tmp/ccs

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal and confirm the CLI is available with claude --version.
  2. Clone the repo into a temp folder: git clone https://github.com/feiskyer/claude-code-settings.git /tmp/ccs
  3. Create the skills directory: mkdir -p ~/.claude/skills
  4. Copy just this skill: cp -r /tmp/ccs/skills/deep-research ~/.claude/skills/
  5. Clean up: rm -rf /tmp/ccs
  6. (Recommended) Register the firecrawl or exa MCP server for better web collection; otherwise it degrades to WebSearch/WebFetch.
  7. For large jobs, make sure parallel (GNU Parallel) and timeout exist — on macOS run brew install parallel coreutils.
  8. Restart Claude Code and ask for a "deep research" task; review the proposed plan and reply "go" to start. Expect noticeable token/API usage from the parallel subprocesses.
View source on GitHub ↗License: MIT